Altair
Skip to content

Trading Analysis Report: AMZN

Analysis date: 2026-10-14 · Model: claude-opus-5-5 · AI-generated, not financial advice.

  • Analysis date: 2026-10-14
  • Rating: Underweight
  • Generated: 2026-10-14 17:29:12
  • TradingAgents 0.6.0: anthropic, deep claude-opus-5-5, quick claude-sonnet-5-5
  • Analysts: market, sentiment, news, fundamentals; research debate rounds 5, risk debate rounds 5
  • Data vendors: core_stock_apis yfinance, technical_indicators yfinance, fundamental_data sec_edgar,yfinance, news_data yfinance, macro_data fred, prediction_markets polymarket

V. Portfolio Manager Decision

Rating: Underweight

Executive Summary: Underweight AMZN: cut to about one third of a standard allocation before the Q3 print. That means selling to about half this week on a schedule and finishing by about Oct 26 and before the first mega-cap peer report, after first checking the 254.08 close, the print date and the implied move. Go to a quarter if the implied move is 10% or more, if your book is concentrated in other mega-cap or AI-capex names, or if you skipped the lookups; hold 40% only if the implied move is 8% or less; if you are flat, do not open a position. Before the print, a daily close below 242.5 takes you straight to target, or the residual to zero if you are already there, because stops will not fill through a 10-15% gap and size is the only real control. After the print, rebuild only against a 10-Q scorecard written in advance, add nothing on a weekly close below 225.82, and use a 3-6 week horizon with a $251 target.

Current Price: 254.08

Price Target: 251.0

Confidence: Medium

Time Horizon: 3-6 weeks

IV. Risk Management Team Decision

Aggressive Analyst

Aggressive Analyst: Since neither of you has spoken yet, I'll take your best arguments in advance.

First, the idea that trimming AMZN is the timid choice. I think it's the aggressive one. The aggressive move is to stop spreading risk evenly and put it where the payoff is best. Going into this print, that isn't a full-size long. The skew is roughly +10.6% if things go well against -15% to -22% if they don't. To break even on those numbers you have to believe the good outcome is 59% to 68% likely. Even if you give the good case the generous July treatment of a +15% gap, you still need better than 55%. Nothing in the data gives me that edge. The daily SuperTrend is down with its stop at 263.49. ADX is 19.69, so there's no trend to lean on. RSI has fallen from 59 to 49 in three sessions, the MACD histogram has fallen three days in a row, and price is below both the 10 EMA and the 50 SMA. Holding full size through a binary event with no edge isn't boldness. It's inertia.

The fundamentals make the skew worse, not better. Revenue is up 19.6% and operating income is up 43%, and I'm not disputing that. But the $5.75 Q2 EPS is mostly non-operating gains. Annualize the $27.5B of operating income, tax it, and you get roughly $90B, which puts the $2.74T market cap near 31x, not the headline 20x. H1 free cash flow was about -$27B, capex was up 68% to $54.2B in Q2 (27% of revenue), and the company raised about $63B externally in H1. Cash fell $23.6B in Q2 and working capital shrank from about $38B to about $8B. This is a long-duration, cash-negative growth story at a premium multiple, in a market pricing a 97% chance of no cuts in 2026 and an 86% chance of another hike. Add oil near $103 and a beta of 1.46, and the setup is poor. Barron's also calls AMZN the most crowded hedge-fund stock, so the people who wanted to own it already do.

To the conservative side, who will ask why I'm not exiting entirely: because the right tail is real and I want to keep it. The weekly and monthly SuperTrends are up, with stops at 225.82 and 177.64. Operating momentum is genuine. The last print gapped +15%. One third of a normal allocation is my call option on that outcome. If the stock closes below 242.5-244, where the 200 SMA and lower Bollinger band sit, the last third goes too. The plan doesn't hope. It has an exit built in.

To the neutral side, who will say it's range-bound, so just hold: a 244 to 264 range with no edge is a reason to reduce size ahead of an event that has historically moved this stock 10% to 15%. Also look at what happened after the July pop. It went from 271.58 to a 284 peak, then fell to 245.96, so buyers of the gap sat on a drawdown of more than 9%. Chasing the print has not been a free lunch here.

The retail bull case is thin. It's one session, 30 posts, and the only bearish voice is a single account. The Anthropic read-through ($16.8B quarterly revenue, +45% QoQ) could be a real tailwind, and one poster says the numbers came after the close. If that pops AMZN toward 258-263, that's the bounce the plan wants to sell into. If it doesn't, the trim has already done its job. Either way, I don't need to be right about the narrative.

There are caveats. The 254.08 close may be an incomplete bar, and StockTwits has price at 243-246. If it's really 244, selling there makes no sense, because you'd be selling into the support the plan uses as its kill line. Verify the close first, and don't chase down to it. The print date isn't confirmed either, so if it's weeks away, there's time to wait for the bounce. And if you're flat, don't initiate.

The real upside of this plan is what it does after the print. Trimming now leaves capital free to rebuild to two thirds on a clean filing and a hold-tier Q3, or to a full allocation on a clean filing and a strong one. Meanwhile it avoids a 15% to 22% gap on a position I can't hedge, since stops don't fill through gaps. I'd rather be the one with dry powder after a miss than the one explaining a stop that never executed.

My call is to trim AMZN to about one third of a standard allocation, ideally into 258-263, and cut the rest on a daily close below 242.5-244. Aggressive Analyst: We all want to trim, and none of us wants to go to zero. The argument is over whether the residual is a quarter, a third or a half of a standard position, and the neutral analyst's own arithmetic says that's worth about 2% of a position either way. So I'll give ground where you're right, then explain why I still land on a third.

Conservative is right that a third of a share position isn't a call option. I used the phrase loosely. It's a small, defined slice of equity exposure. If someone wants the literal option, they should price calls once they've seen the implied move, which we don't have.

You're both right that the sale shouldn't depend on a bounce into 258-263. The plan says "ideally," but "don't chase down" reads like a gate, and a gate leaves you full size if the stock slides first. So sell the first tranche once the close is verified, treat 258-263 as a reason to sell faster, and finish a couple of sessions before the print.

Neutral is also right that 31x is the harsh end. Forward EPS of $10.45 on about 10.8 billion shares is roughly $113B of earnings, so the honest number is probably high 20s. That doesn't change the thesis. A high-20s multiple on negative free cash flow, with capex at 27% of revenue, still depends on things our tools can't see. And I agree on July: holders got paid, and the gap chasers got hurt. The rebuild tree is built around not chasing.

Now the pushback. Conservative, you said a close-based exit can't fire until the gap has already happened. That's true only for the gap session. If the late-October pattern holds, there are roughly ten to twelve sessions before the print, and a close under the 200 SMA and lower band in that stretch is exactly when you'd want the rest gone. The kill line was never meant to protect across the gap. Size does that, and we agree. I also never used the weekly SuperTrend at 225.82 as protection. It only gates when I'm allowed to add back. And you wrote "one big pop in four" and then listed October 2025 at +9.6% in the same paragraph. It's two of four, which isn't an edge either.

Neutral, you're right that 22% is February's low projected onto today's price. But 11% is a floor, not a ceiling. The tool starts February at 222.69, the first heavy-volume close, and doesn't show what came before it. If any of the selling happened ahead of the print, 11% is understated. Four events also tell you the worst case so far, not the worst case. This is a 1.46 beta, the most crowded hedge-fund stock, with a macro backdrop that offers no multiple tailwind.

Your own math then says expected value is roughly zero. If so, what is the extra sixth of a position being paid for? It isn't edge. The ±7% you cite is one StockTwits poster, not the options market, and the realized moves in the four windows averaged somewhere between 8% and 10% depending on how you count February. Your rule says go toward a third as you approach 10%, and the history points that way.

The kill line also isn't selling the range low. The intraday lows were 244.30 and 244.73, and the lowest close was 245.96. The line is a close below 242.5-244, which is under the 200 SMA, the lower band and both prior lows. That's the range failing. Waiting for a second close is defensible, but it only costs money, and it's the least important disagreement here.

On the technicals, I agree they don't say sell. They say nothing, and that was my point. Nothing is a poor basis for full size going into an event that has moved this stock 10-15%.

Conservative, your loss-budget framework is the right one. A third at a 22% gap is about 7% of a standard position, and you said to ask whether you'd shrug at that. If you would, the third stands. If the book can't take it, hold less. We can't answer that from here, since portfolio context wasn't provided.

When edge is zero, what you control is variance, and what you want to preserve is the ability to act after you know more. The filing will show the source of the $35B-plus of non-operating income, the debt composition and the other investing outflows. Selling now isn't a forecast of a miss. It's declining to carry full variance on a coin flip, at almost no cost in expectation, and keeping capital free to rebuild with real information. Neutral says rebuilding after a gap costs the gap, but that's the mirror image of the trim saving the gap. The cost is symmetric, and the third that stays keeps me from being shut out.

Verify the price first. Several StockTwits posts were stamped after the regular close, so the 243-246 quotes could be after-hours rather than a different close. Nothing in the data confirms that. If the true price is about 245, the skew from there is roughly +14.7% to a 281 target against -15% to -19%. That's close to balanced, and selling into support makes no sense. Wait for a close.

So the plan is to verify the close, trim on a schedule to a third, write the kill line down beforehand, do nothing on gap day, and rebuild only after reading the filing. If the implied move comes in at 10% or more, I'd accept going toward a quarter, because variance is the thing I'm sizing. I wouldn't go to zero, and I wouldn't hold half through an event I can't hedge and can't size from here. Aggressive Analyst: I'll start with the corrections, because several of them are right.

Neutral is right that the kill line has to expire at the print. A gap from 254 to 220 closes below 242.5 on the gap day, so a rule that says "cut the rest on a close below" and a rule that says "do nothing on gap day" can't both stand. The close-below rule is a pre-print rule. After the print, the residual gets decided from the filing and a session or two of digestion, with 225.82 as the line where adds stop and the remainder gets reassessed. I'm folding that into the plan.

Neutral also caught something sharper than my after-hours guess. The bar's low is 252.66, and a regular-session price of 244 would sit about 3.5% below that. So either the 243-246 quotes are after-hours with an unexplained move, or the bar was snapshotted early and the real close is near 244. Those two readings lead to different branches of the plan, and checking which one is true is a thirty-second job.

Conservative is right that size shouldn't depend on where the tape sits. The tape only affects execution. That's why I'd drop the bounce target as a gate and sell on a schedule, with strength into 258-263 as a reason to sell faster.

I'll also stop leaning on the 22%. It's February's low projected onto a higher price. At a 15% stress and +10.6% on the good side, breakeven is about 59%. Use July's +15% on the good side and it's 50%. That's zero edge, which is my point. The case for trimming never needed negative edge. It needs a coin flip with fat tails that I can't hedge.

Now, why a third and not 40%. Neutral's own ladder says 40% at an 8-9% implied move and a third at 10% or more. We have no implied move, so the default has to come from the tape. The two windows where we can see the gap cleanly are October 2025 at +9.6% (222.86 to 244.22) and July at +15.3% (235.50 to 271.58). They average about 12.4%. February starts at its first heavy close, so its true size is unknown and the 11% is a floor. On Neutral's rule, a 12% average points to a third.

The positive-drift argument is real, but over a three-week window it's a few tenths of a percent on the marginal sixth, which is noise. Friction on that sixth is a penny of spread plus a tax bill that's mostly owed either way. When both sides of a marginal sixth are that small, I'd rather it not sit in a zero-edge position, because it uses up risk budget that could go somewhere with an edge.

Conservative, you keep saying the smaller residual is the safe resolution of uncertainty. Uncertainty never goes away, so that rule has no stopping point. It would have you at a quarter this quarter and at zero by next print. Neutral made the same point: ignorance alone would have you sell before every print. What actually justifies a smaller position is unhedgeable variance, and variance is measurable. Your own formula gives a third at a 5% tolerance and a 15% stress. That's the whole disagreement, a 5% tolerance versus a 4% one. Both are defensible. But a third caps the bad case at 5% of a standard position against 15% at full size, so the trim removes two thirds of the variance. Everything from here on is arguing over the last point or two.

On "a mistake I can recover from," I accept it as a heuristic. But a missed pop on a stock with accelerating revenue, 43% operating income growth and weekly and monthly trends still up is a real cost. The third I keep is what stops me from being shut out. Your own claim that you can rebuild because the trends are up also concedes the right tail is live.

On the kill line, I'll take the first close inside a week of the print, as Neutral offered. Earlier than that I'd also take the first close, as you want. It's the least important disagreement, and it only costs whipsaw.

On the rebuild, I'll take Neutral's version. Two thirds comes only after I've read the 10-Q for the source of the non-operating gain, the debt mix and the other investing outflows. A full allocation comes only if price holds above the gap and the 50 SMA after that. Conservative's veto stays: if the filing looks wrong, I don't add. July is why this is cheap insurance. Gap buyers at 271.58 were down 9.4% within weeks, so paying for confirmation costs less than chasing.

Here's the version I'd stand behind. Verify the close. Sell the first tranche this week to about half, as Conservative's waypoint and Neutral's two-tranche structure both suggest. Sell the second a couple of sessions before the confirmed print, down to a third. If someone has the implied move, I'll move with it. At 7% or less I'd accept Neutral's 40%. At 10% or more, or with correlated AI-capex names in the book, I'd go to a quarter. The kill line applies only before the print. Don't initiate if flat, and sell nothing if you're already at or below a third.

The three of us are now within a quarter to 40% of each other, and a third sits in the middle. I'd rather go into this print with a position I can ignore than defend a bigger one against a gap I can't hedge. It's still a Sell, trimmed to a third. Aggressive Analyst: Conservative landed a real hit, so I'll start there. I used the 12.4% average to argue against 40%, then said I'd drop to a quarter at a 10% implied move, and those don't fit together. Neutral is right that 12.4 is the mean of the only two up-gaps where the tape shows the prior close. It's the top of what we can see, not the center. Count February's reaction day at about -5.6% and leave April unsized, and the visible moves run 8 to 12 with a middle near 10. That's the boundary between Neutral's 40% rung and the third rung, so a third is the right default without the options number. A quarter needs a confirmed implied move of 10% or more, or a book that already stacks other AI-capex names. My argument last round was just too strong.

That leaves us closer than the tone suggests. Neutral retired 40% as a default and said to stand at a third without the implied move, which is my number. The remaining gap is Conservative's quarter, and it rests on assuming the book is full of correlated mega-cap names because we can't prove it isn't. That's the same "unknowns argue for smaller" move aimed at a new unknown. Conservative gave the stopping point earlier: three lookups, then the ladder. The book is a fourth lookup, and it belongs to whoever owns the position. The loss-budget check handles it. The residual times 15% has to fit inside what you'd shrug at, and if the book is stacked it won't, so you go lower. We can't run that for the caller, so the default should work for a standard book, and that's a third.

On "when edge is zero the tie goes to the smaller loss," a tie rule like that ends at zero, and something else has to stop it. The loss budget stops it. Mine is 5 points of a standard position at a 15% stress, which gives a third, and Conservative's is 4, which gives a quarter. Both are defensible, but neither is the unrecoverable mistake. A third at a 15% gap costs 5 points and a quarter costs 3.75. That's 1.25 points apart, and on a +12% print it's 1 point apart the other way. The unrecoverable mistake was full size, and all three of us have already sold that.

Every ladder on the table also passes through half this week. A quarter, a third and 40% all sit below it. So tranche one isn't in dispute, and the fraction only gets decided at tranche two, a couple of sessions before the print. By then we should have the true close, the date, the implied move and possibly MSFT and GOOGL capex commentary. Two tranches buy that information. That's the reason not to jump to a quarter today, and I'd happily put tranche two after the peers report if they land first.

On the post-print rule I'll take Neutral's split over Conservative's all-out. Conservative is right that digesting a down gap can mean riding a second leg, and February shows it. But scale February to today's price and the path is roughly 240 on the reaction day and about 227 at the low, almost exactly the weekly stop at 225.82. Neutral's rule sells half the residual at the next open after a close below 242.5, and the rest on a close under 225.82 or a second close under 242.5. Conservative's sells everything near 240. On that path the difference is under a point of a standard position. If the flush rebounds like 9/16 and 9/28 did, the all-out rule sells the low on the whole residual. It's a coin flip on small stakes, so write it down beforehand and follow it. If the gap is up or holds above 242.5, do nothing until the 10-Q is read.

On the price, Neutral's catch is sharp. A 252.66 session low can't coexist with a 244 regular-session quote, so either those quotes are after-hours or the bar was snapshotted early. I agree with Conservative that both branches are one retail-sourced session and neither should change the size. They change execution. If it's really 244, we're sitting on the kill line, so the schedule still runs and we sell the week's bounces, not an intraday flush, and the line handles a break. Selling strength into 258-263 is a reason to go faster, not a gate.

The tax point is fair. In a taxable account with a low basis it's an input to the loss budget, and I'd sell the highest-basis lots first. I also agree with Neutral that the note should be reframed. With a high-20s multiple and something like +10 to +15 against -11 to -15, this is a coin flip with fat tails, not a bearish thesis. The 31x and the -22% were my numbers, and I don't want them biasing the rebuild, which should run off the filing.

This is still the aggressive position. Don't spend risk budget on a zero-edge coin flip you can't hedge. Keep a slice for a right tail that has paid in two of four windows, with operating income up 43% and the weekly and monthly trends intact. Keep the powder for after the 10-Q, when information is highest.

The plan is to verify the close, the date and the implied move. Sell to about half this week. Then go to a third by the print, to 40% only at 8% implied or less, and to a quarter at 10% or more or with a stacked book. A first close below the 200 SMA and lower band before the print takes the rest. After the print, apply Neutral's split and add nothing until the 10-Q is read. Rebuild to two thirds only on a clean filing, to full only if price holds above the gap and the 50 SMA, and add nothing on a weekly close below 225.82. Don't initiate if flat, and if you're already at or below a third, there's nothing to sell. It's a Sell, trimmed to a third. Aggressive Analyst: Neutral, your hit on the rebuild tree is the one that lands, and it matters for my own argument. As written, two thirds needs a clean filing and a hold-tier quarter, and a miss isn't hold-tier. Full needs price back above the gap and the 50 SMA near 258, which after a gap to 220 is a 17% climb. So the only branch that ever spends the powder is a clean up-gap, where I'd be buying above 271, the chase I keep saying I won't make. Without a down-gap branch, my dry-powder line is rhetoric.

I'm adding your branch. A down gap with a clean filing means the drop is explained by multiple and flow, not by operating income or operating cash flow breaking. It also means the source of the non-operating gain is understood and the debt mix is fine. In that case, two thirds comes back on a weekly close above 225.82, and full comes back on a reclaim of the 200 SMA near 242.5. That is where the trim pays for itself, because you're buying back roughly 12% below where you sold. A weekly close below 225.82 flips the weekly tier, and nobody adds into that.

Conservative, you win on the peers. I asked what decision the peer read changes and I can't find one. Nobody is arguing the read-through earns a bigger residual, and a peer capex print is another sympathy gap on the same story, which is the variance I'm trying not to carry. If the peers report first, finish before them. If they land a day or two ahead of ours, the two rules coincide. I'll also take skipped homework, framed as Neutral did, as a precondition for tranche two and not a penalty. If you've had two weeks and haven't spent thirty seconds on the implied move, you don't get the default, and you land at a quarter. That costs me a point and forces the lookup, which is worth more than the point.

I'm not moving on the heavier left tail. Neutral is right that the cash items were on the table in July. Q2 capex was up 68% and quarterly free cash flow was about negative $9B, and the stock still went from 235.50 to 271.58. Whatever the cause, it then took 13.4% off the high close by mid-September, so some of that worry has already been paid in price. The no-cuts pricing and the crowding headline are known too. What nobody can see from here is Q3's cash line, but they also can't see AWS growth, backlog or what Anthropic's numbers mean for Amazon. Unseen on both sides is symmetric.

Crowding is a fair reason the left tail is fat, so it goes in the note. Operating income up 43% with the weekly and monthly trends intact is a fair reason the right tail is fat, so that goes in too. The note should say coin flip with fat tails on both sides, with the cash items listed as rebuild checks, which is where you wanted them anyway.

On the index fund point I'm with Neutral. If owning an index fund counts as a stacked book, the flag fires for everyone and the third never applies. Count AMZN's look-through weight in the loss budget, and move to a quarter for a concentrated overlay of mega-cap or AI-capex names, because that's where losses really stack.

You're also right that the written rebuild, not the residual, is what keeps me from being shut out. I overclaimed. The residual is position P&L. Sized as P&L, a third is five points of a standard position at a 15% stress and a quarter is under four. For that you keep a stake in a right tail that has paid in two of four windows.

Two cleanups for the trader. First, pin the kill line to one number, a close below 242.5. A zone of 242.5 to 244 is ambiguous on the day it matters, and 244 is where the range held twice. If the bar was caught early and the real close is near 244, we're already at the line, so sell nothing into the open and wait for the close. Second, the 10-Q may lag the release by a day or more, so anything the release doesn't answer waits. The written post-print rule covers that gap in time.

Here is where I land, which is Neutral's list plus the down-gap branch:

  • Before the sale: Verify the close, the print date and the implied move.
  • Tranche one: Sell to about half this week without waiting for 258-263, and treat strength there as a reason to go faster.
  • Tranche two: Finish before the first peer report and no later than a couple of sessions before our print. Default to a third. Go to 40% only at an implied move of 8% or less. Go to a quarter at 10% or more, with a concentrated overlay, or if the lookup was skipped.
  • Kill line: A close below 242.5 takes the rest, and the rule expires at the print.
  • After the print: An up gap or a hold above 242.5 means leave it alone until the filing is read. A reaction-day close below 242.5 sells half of what's left the next session. The rest goes on a close below 225.82 or a second close below 242.5.
  • Rebuild: Two thirds only on a clean filing, with the new down-gap branch. Full only on the price conditions above. No adds on a weekly close below 225.82.
  • Edge cases: If you're flat, don't initiate. If you're already at or below target, you're done.

Three of us started with different views on how much to hold through this print and ended a rung apart, with the disagreement now worth a point or two of a standard position. The thing worth being aggressive about is not spending risk budget on a zero-edge coin flip you can't hedge, and keeping the capacity to buy something with real edge once the filing has been read. It's still a Sell, trimmed to a third.